Artificial intelligence always manages to attract attention with new headlines in image processing. Here a new denoiser, there a deblur or even a low-light enhancer. A Google research team has now gone one step further and combined many such tools into a single Transformer model. This is called MAXIM and masters the following image enhancements "in one fell swoop":
- Low-light enhancement
- Image retouching
- Dehazing indoors
- Dehazing outdoors
- Image deraining
- Image deblurring
The code is available via Github for anyone to try out, but at the same time the team makes it clear that it is not an official Google product. However, since CoLab notebooks are also linked, you can immediately play around with the model if you have some Python experience. And via Hugging Face Spaces you can even play around with your own images directly in the browser. The same even works via a MAXIM demo site hosted by Google.
MAXIM handles various image enhancements in an AI model
Since the model has been ported to Tensorflow in the meantime, we may be witnessing the emergence of a universal filter for image enhancement that will be found in various open source projects very soon. As already shown in Stable Diffusion, the open source community is currently far faster in implementing current AI shaping results than industry giants such as Adobe or Apple.
And the pace of development remains staggering. Who would have believed two years ago that it would really be possible to "deblur" images? Since Google certainly has its own in mind with this tool, we will certainly find something similar in the Android camera app in the near future.